3369: Precisely Explains Why AI Fails Without Data Integrity

[00:00:04] How can businesses trust AI when they don't fully trust their own data? Well, in today's episode, I'm going to be sitting down with Andy Bell from Precisely. And together, we're going to unpack exactly why data integrity is often considered the missing link in so many different AI strategies right now.

[00:00:24] And Andy will explain how third-party data enrichment and better governance could possibly turn incomplete or reliable information into a solid foundation for AI that actually delivers results. What would that be? Saving millions in delivery logistics to building risk models that protect entire communities? We're going to explore how enriched, trustworthy data can power that AI that we keep hearing about.

[00:00:52] And as a result, earn confidence, not suspicion. So with a more optimistic look today about AI, I'm going to officially introduce you to today's guest. So a massive warm welcome to the show. Can you tell everyone listening a little about who you are and what you do? Yeah. Hi there, Neil. My name's Andy Bell. I head up data product management at a company called Precisely. And effectively, I look after actual data.

[00:01:22] So the ones and zeros, we build data products for our customers that use it on a global scale, across multiple use cases. And that is an important differentiator for Precisely, who are a leader in data integrity. They're all around building accuracy, consistency, and context in the data that our customers are using.

[00:01:44] And being able to add additional sort of third-party external data sets to the work that we do with those customers is a critical part of our business. And I've been in the data business for 30 years, I think. So we've seen a lot of change over that time. Probably none more so than what we're seeing right now. So, yeah, interesting times for us. Yeah, we're in such an exciting time at the moment. But for both of us, we've seen so much excitement throughout the years as well,

[00:02:13] whether it be the arrival of the internet, computers in offices, and mobile app stores, the cloud. And I think very often people have said that AI is very much like the iPhone moment. But I would argue it almost feels like the App Store moment. It begins with all these little gimmicks. We're turning ourselves into action figures where mobile apps we would turn into chainsaws and a pint of beer. But we're now getting to that maturity phase.

[00:02:38] And there's been a lot of AI projects that have been stuck in pilot phase, a lot of questions around ROI, so many different parallels with before. But I'm curious, there is so many companies racing to implement AI and be innovative with new use cases. From what you're seeing, how successful has this been from your perspective? Yeah, I mean, AI can bring a lot of value to organizations. It is interesting how it's sort of come through consumer, and that's where it's very quickly sped up.

[00:03:07] And then businesses go, how can we use this? I think what we've seen is a lot of companies jumping in and then suddenly finding the flaws that they do see within AI around. If you don't have really good data or you can't trust that data, we can see the hallucinations, we can see the bias and the unreliable outcomes. And the problem with something like that happening is people then lose confidence.

[00:03:36] And I think that is a problem with where we're at at the moment is how do we gain that confidence in AI giving us consistent and reliable insights and information and direction and those types of things. And I think it's so critical that being a data person, having the right data and having trust in that data is just really, really critical.

[00:04:01] And we do a lot of surveys, as do many companies, and we're seeing a very small percentage, around 12% of the businesses we work with, having any confidence in the data that they're using within AI. So I don't think they feel necessarily ready for it, but it's there and people want to be using it. Yeah, such a great point mentioning data, because I think when AI first arrived, I think generative AI were coming up to three years now.

[00:04:29] And I think everyone just jumped headfirst in. Everyone wanted to be on that bandwagon, but nobody was talking about, hey, what about my data silos? What about my data integrity? And these things are at the heart of what AI and machine learning need. So just to hammer home this point, how important is that role of data integrity playing in AI strategies? And what are the implications of leveraging unreliable data for these strategies? I suspect it's the age-old problem of garbage in, garbage out. But what are you seeing here?

[00:04:58] No, it absolutely is. It goes back to what I was saying around the challenges that are coming out, and you don't have reliability around the outcomes. And yeah, I think it's just, it's so important. And if we start going into what we're seeing is the, you know, the move to angentic AI, which is it's autonomous. You're not really understanding. Well, if you really aren't getting, if you haven't got data integrity, you haven't got that trust in the data, can you trust the automatic responses?

[00:05:26] And I kind of flip it back to what we expect from businesses as consumers. We expect consistency, reliability, correct information, and those types of things. And I think that if that's not happening, that can really undermine one confidence, but also can be very detrimental to a business's brand.

[00:05:47] If they're using systems and using data that's not correct, and people are not having the experience they would expect to have with an organization that they may interact with on a regular basis. A hundred percent. I mean, I've been to so many tech conferences this year, and predictably, every single one is talking about agentic AI.

[00:06:07] And there is that cautious IOM ex-IT guy here thinking, you do realize that putting swarms and swarms of AI, agentic AI agents out there that are not fully trusted that you don't know that much about, you could be causing a few problems down the line. It's much bigger than shadow IT and BYOD, isn't it? Yeah, it absolutely is. And not only that, what you don't want to get into a situation is you might have a customer goes, why have you made that decision? They go, well, no idea.

[00:06:36] It's this thing over here that's made the decision for me. You've got to be able to have traceability around if a decision is made, particularly in regulated industries, have you made the decision on the right information? And what's influencing that decision? And how do you know which bit of it is wrong so you can correct that for that consumer experience? Yeah, it's going to be tricky. Yeah. And who's accountable when one of these agents makes a mistake?

[00:07:05] Who's in charge of that particular bot? Well, yeah. And it almost gets to who owns the IP, ultimately. Is it the algorithm or is it the data or is it the organization processing that information? I don't think we've really tested that yet. But it is. You've got to think where does accountability and governance come into that process in particular? Yeah.

[00:07:30] So as we follow that big tech mantra of move fast and break things and businesses use AI more and more, how can they better ensure that they make strong decisions with the right insights that are generated? Is that a way they can enrich their insights and ensure that they navigate around those possible mistakes that we just mentioned? Yeah. I mean, this is a really interesting area because it's where me as a business, my part of the business kind of comes into that.

[00:07:59] So lots of organizations have their own internal data. They have their customer data. That's a very particular view of the world. That's their customers. They will attract a certain type of customer and they will only have so much information that it might be incomplete. So one of the things we see is organizations are coming to people like Precisely to get additional enrichment data.

[00:08:24] And that's really attribution that we can add to a customer record, either to get more complete insights on data that they've already got. That might be age, it might be location, those types of things. Or it might be going, what additional information can I obtain about that customer? So they could be going, what type of neighborhood do they live in? Do we understand what type of natural hazards that the property they live in is exposed to?

[00:08:53] That sort of external information to give greater context. And really, when you think about that, it's also about getting that picture of the market as well. It's right. I go back to that thought of these are my customers, but this is the world. This is the market. I don't know what those other people, what those other situations look like. What's my opportunity and what can I go to look for?

[00:09:18] So we kind of see both of those aspects that customers come to look for sort of external and additive data sets to what we're doing. And we're seeing that increasing. Again, we go back to our surveys. We've got around about a third of respondents are now looking for that external data enrichment. And I think that's going to increase with AI in particular. And we've spoken a lot around data today.

[00:09:43] What would you say are the biggest implications of businesses data not being contextual enough? And any tips or advice on how to navigate around a hurdle like this? Yeah, I think it is. You've got to think around how they understand relationships, how they can understand the relationship to location. So one of the aspects that I didn't introduce is the data that I look after all has a location characteristic about it.

[00:10:11] And that in particular is a really interesting impact on what a customer thinks about, what a business thinks about, and how that influences the decisions they're making. And without that context, you can't really drive what you're trying to do.

[00:10:29] So you think about something like supply chain management, a really good sort of insight to that recently, or I think it was last year when, I don't know if you remember that ship hitting the bridge in Baltimore. Yeah. A major piece of transport infrastructure taken out. We have loads of customers coming to us and going, what is the impact on our supply chain? Not only, but also what's the impact on our businesses that we're working with? Supply chain.

[00:10:59] And understanding and getting that sort of broader context around an event and how that impacts what you're doing. But it's also things like, you look at marketing, making sure you've got the right address for a start. But also, when you think about things like click and collect, is are you actually providing them a collect location that's convenient to that customer? It's easy to access, those types of things. So it's kind of bringing in that sort of broader context.

[00:11:28] But I know my customer and I know my location says, how do you connect all that information together to get that greater insight to how you're performing and how you can provide better experience? And we've seen over the last few years, big tech companies have been hoovering up all online content to train their ML models and even rumored to be running out of training data. And this got me thinking about how businesses could better use third party data.

[00:11:54] So are there any real world examples where third party data might help deliver tangible business outcomes? Any examples there? Oh, absolutely. I mean, more than I can go through at the moment, I'll give a couple of really good ones. We're working with a major delivery company at the moment. So Food Delivery, the type of DoorDash Deliveroo, Uber Eats, that type of organization. And they were a challenge they had, not delivering first time, having incorrect addresses.

[00:12:23] They're delivering people taking far too long to find the address and also taking inconvenient routes. So we were able to sort of bring additional data sets to them and capabilities around how do we get them and help them deliver first time, right time to the right address. And that was critical because they were losing to competition. They were losing customers and it was costing them a lot of money.

[00:12:53] Once we'd implemented that one, apparently their delivery people got bigger tips. They got better customer satisfaction and saved $65 million in lost deliveries and late deliveries. So it has really tangible impact on the business bottom line. We work with San Bernardino. Obviously, they have lots of wildfires in California, unfortunately.

[00:13:18] We run a wildfire risk model and they really wanted to understand, OK, we know where the fires have been, but where's the risk, the future risk for us? We're able to bring our wildfire risk models, property information and disinformation to help them understand where that risk is and what the potential impacts and what they would need to do to think about how they might evacuate an area. Because you've got to think around when you're looking at something like that is who do I, where do I prioritize?

[00:13:46] Is it the old people's home? Is it the hospital? All those types of things. So it goes from delivering your food to saving property, saving lives. Yeah, it's really, really varied set of use cases that we help our customers address. Wow. I suspect you've just set off so many light bulb moments around the world as people are listening, thinking about their own data, getting their own data in order,

[00:14:10] and never even thought about how third-party data might be able to provide a measurable impact, ROI, on their big tech projects. So I'm curious, for anybody listening and maybe they want to pursue an avenue like this, what are some of the challenges that organizations typically face in managing and using third-party data effectively? Yeah, absolutely.

[00:14:31] And I spend a lot of time talking about this because you're going to go – because from our point of view, sometimes it's like you just think, well, we've just got to convince the customer this is the best data. You go through an evaluation, proof of concept. You go through the legal contracting, you contract it, and then we deliver it. And everybody thinks that's it. It stops there. Absolutely not. So the customer then thinks, well, how do I onboard it? How do I keep it up to date?

[00:14:58] How do I get into my models and train my models? There's a huge catalog of things. And quite often, that post-sale onboarding of third-party data can be more expensive than actually acquiring the data to begin with. And it's a really important part of what we're focusing on is how do we reduce that time to getting value out of those data sets.

[00:15:26] So the customers have got to look and think about the questions they need to ask their suppliers, how they maintain and update data. But it's also – and there's a couple of areas that we're looking at improving that situation. One, how do we deliver it in a way that your customers want? So think about how long we've been. It wasn't that long ago feels that we were just slapping data on a DVD, pop it in the post, and off we go. No.

[00:15:54] Digital downloads, everybody expects that. But they also want an API. They want a managed service. They want it delivered into Snowflake. They want it delivered into Databricks. They want some analytics. We focus on every single aspect of that and deliver on every part of that. And that is – you just think about delivering into Snowflake. One, you're delivering directly into our environment. But once you have that set up, it's just automatically updating.

[00:16:21] Once we put an update in place, everything's just flowing through. So they're not going, right, I've got to wait for the email. I'll then download it off this slide. I've got to unpack it all. And then I've got all those things. So that's a real reduction in time and effort. And then the other part of it is how you connect data together. So you can think about it by data from all these different providers. And particularly when you're talking about location, that can get really, really complex.

[00:16:48] So one of the things that we've been doing is, one, connecting all of our different data sets together. We use unique and persistent IDs. We focus on applying that to an address. So all the customer needs to think about is, I've got an address. They're very used to doing an address validation and checking. And then we go, you can then add all of this data. You don't need to do a whole load of complex processing to access multiple different data sets from, what's the property attributes of that business?

[00:17:17] Is it in a wildfire risk area? What neighborhood is it in? What type of people live in that area can all be done in one simple query? What we've now taken that to is, you think about, okay, I might access multiple data sets from multiple different sources.

[00:17:35] We have developed a program called Datalink, which is going out and connecting to other ID systems that other companies and organizations run. And this is having a really interesting impact within the market. So we're already connecting with Overture Maps.

[00:17:58] That's the mapping data program that's been run by Meta, Microsoft, AWS, and TomTom. So we're linking into that. And then we're also linking into B&B's DUNS number. So a customer can think, I've got a DUNS number. I want to know more about this business. They can come into our API, enter that DUNS number, and access a correct address. They can access what other businesses are co-located.

[00:18:28] Is it in a flood zone? All of those types of things just by having that single ID. And we kind of feel that that kind of gets into that sort of universal data adapter kind of capability. But we think that's going to bring significant benefits to our customers and also the customers of the partners that we're bringing into that. So people like Overture, D&B. I think one of the things I think about is, as a business, we consume huge amounts of supplier data.

[00:18:57] We know how painful it is. And when I talk to my engineering is, think about how painful it is for you. Don't make it that painful for our customers. That's what we've got to think about. Yeah. Love it. And as we do see more maturity from AI, and we finally see things coming out of those AI pilot phases. It's fine to move beyond that low-hanging fruit, solving real problems. What is your big focus for the rest of this year and maybe early next year?

[00:19:26] And what is it that excites you about all the tech trends and opportunities ahead? No, no, absolutely. I think there's two things. One is, I was talking about data link, is how do we expand that to other companies and bring additional data sets and make that easier? Because that then leads into the second part is, how do we put AI capabilities over the top of all of that portfolio of data and the data link

[00:19:54] and around how a customer can interact with third-party external data just during a natural language search? And I think some of the things we're seeing around MCP sitting in between, say, something like Claude and our API, I just think that's really, really exciting because it's like, and why it excites me, it's not because it's a bit of tech,

[00:20:16] is it makes our end users' lives so much easier and opens up the market to accessing data that they may have felt, oh, it's just too complex. I don't want to think about that. If you think about location data, I don't want to think about pointing polygons and mirrors. No, you just have to ask the question, where can I find the best locations for the shops? And I just want to kind of look at it and think about, I want to be this kind of thing, and I want to be these types of people.

[00:20:46] And you just, right, I think that's a really exciting, much more interesting way to be able to start thinking about how do I solve my business problem. Brilliant. I think that's a great moment to end our conversation on today. But before I let you go, I always like to have a little fun with my guests and also ask them to share something a little more than just their insights on business and technology. Now, before we started recording today, you were sharing a story where you appeared on Six Music.

[00:21:14] So it seems only natural that I should ask you, what is a song that has inspired you or means something to you that we can add to our Spotify playlist? What would you like to add and why? Yeah, well, it was quite challenging because I did look at the list and went, well, it goes from Eminem to Edith Peer. Blimey. But brilliant question.

[00:21:33] And I love things like this because if anybody ever follows what I post out on LinkedIn, there's a lot of music and geography and also talking about my son, my youngest and his band and his music career and everything. So music has been an important part of my life.

[00:21:48] And I think what I've chosen, I'll get to it in a moment, is thinking it in the context of when I think about a career, there's always been key moments in your life where either you've had a job change, promotion, or you've released a brand new product and you've got your first big customer. These things are key parts of your career. And I think about music in that way. So I've gone right back to the beginning and gone, this song as an 11-year-old that I heard. Wow, that's brilliant.

[00:22:18] And I love it. And it's kind of guided my musical taste for the rest of my life. And it's the jam going underground. Oh, what a great trip. You had me on tenterhooks there. I thought, what's going on with this, man? We'll get that added to our Spotify playlist. And you've got me in a generous mood. I'm going to try and give you some cool dad points now. Can we give your son's band a shout out? What's your band's name? Well, they were called Casino Havana. Unfortunately, they're kind of laid low at the moment and they're probably going to come out with a different new name.

[00:22:47] But absolutely, go listen to them. They're good. Awesome. A quick shout out there. And for anyone listening who just wants to find out more information about Precisely, the work you're doing, everything we talked about today, maybe even connect with you on LinkedIn. Where would you like to point with everyone? Absolutely. Well, we can find lots out at Precisely.com. In particular, if you go to Precisely.com slash Trust25, that's our very recent event, customer event. It's all online.

[00:23:15] So you'll hear a lot about AI readiness, but also dig into, you'll hear from DoorDash around the delivery use case that we've been solving with them. But absolutely, they can find me on LinkedIn as well. And I do post occasionally, regularly. Not occasionally, regularly. Well, I will connect with you on LinkedIn personally. I want to find out more about some of these musical references you put on there. I'll put links to everything for people listening to find it as well. I've had a lot of fun with you today.

[00:23:44] Just a big thank you for stopping by. Talking about how to unlock so much ROI and measurable impact from technology, but in a language everyone can understand. Thanks for joining me. Lovely, Neil. Thanks so much. So as AI continues this rapid climb up the business agenda, one truth keeps resurfacing. Quality in means quality out, and the best way of avoiding garbage in, garbage out.

[00:24:12] Whether it's through robust governance or fresh context from third-party sources, companies must address that data integrity problem head on, especially if they want AI to become a growth engine rather than a liability further on down the road. But what steps are you taking to build trust in your own AI-ready data? Love to hear your thoughts on this one. Join the conversation.

[00:24:37] Email me, techblogwriteroutlook.com, LinkedIn, just at Neil C. Hughes. Let me know how you're closing the gaps between raw information and reliable AI insights. But that's it for today. I'll be back again very soon with another conversation. So a big thank you to Andy for engaging me in a little bit of fun and mischief today. I was having a bit of a cheeky day. And also a big thank you to each and every one of you for tuning in. Speak with you all again soon. Bye for now.

[00:25:07] Bye for now. Bye for now.